claude-review

Files (2)

• SKILL.md
• claude_review.py · scripts/

SKILL.md (first lines)

# Claude Review

Use Claude as an independent reviewer for Shannon issue designs, code diffs,
experiment results, and architectural decisions.

The review is advisory. Codex remains responsible for deciding whether Claude's
feedback is correct, acting on valid findings, and explaining any rejected
findings.

Do not rely on Claude session memory alone. Session IDs are useful for grouping
related reviews and may help continuity, but every important review prompt must
include the essential issue, diff, log, and question context explicitly.

## When This Skill Applies

Use this skill when:

- the user asks for "claude review", "ask Claude", or similar;
- an experiment is large, risky, or touches the Nushell/Reedline fork surface;
- an experiment has already failed once and the next design needs review;
- a code change affects architecture, shell state synchronization, process
  handling, signal handling, config loading, terminal behavior, or persistent
  state;
- before closing an issue after a complex series of experiments.

## Review Timeout

Every Claude review has a hard timeout of 15 minutes.

Use the helper script in this skill instead of calling `claude` directly. The
script enforces the timeout with Python's subprocess API and terminates Claude's
process group on timeout because macOS does not provide a portable `timeout`
command by default.

## Files and Logs

All review state and output lives under:

```text
logs/claude-review/
```

The helper script maintains:

| Path                                         | Purpose                         |
| -------------------------------------------- | ------------------------------- |
| `logs/claude-review/current-session-id`      | reusable Claude session UUID    |
| `logs/claude-review/<timestamp>-prompt.md`   | exact prompt sent to Claude     |
| `logs/claude-review/<timestamp>-stdout.json` | raw Claude JSON output          |
| `logs/claude-review/<timestamp>-stderr.log`  | Claude stderr / timeout details |

These logs are intentionally under `logs/`, which is gitignored. They may still
contain sensitive diffs or local paths. Keep context narrow and do not write
secrets into review prompts.

## Process

1. **Prepare the review context.** Include only the context Claude needs. Prefer
   concrete artifacts:
   - the user's current request;
   - the issue file and experiment section;
   - relevant `git diff` or `git diff --staged`;
   - test commands and observed outputs;
   - screenshots/log paths when those are part of the evidence.

2. **Ask specific questions.** The prompt should tell Claude what kind of review
   to perform. Examples:
   - "Audit this experiment design for contradictions, missing verification, and
     architectural risks."
   - "Review this diff for correctness, regressions, and missing tests."
   - "Given these logs, identify what failed and what experiment should come
     next."

3. **Run the helper script.**

   ```bash
   python3 skills/claude-review/scripts/claude_review.py \
     --context issues/0776-pdf-not-loading/README.md \
     "Audit Experiment 4. Focus on whether the scope is right and whether the verification would prove the result."
   ```

   For a staged diff review:

   ```bash
   git diff --staged | python3 skills/claude-review/scripts/claude_review.py \
     "Review this staged diff for bugs, regressions, and missing tests."
   ```

   For a fresh thread:

   ```bash
   python3 skills/claude-review/scripts/claude_review.py --new-session \
     "Start a new review thread. Reply with the review session status."
   ```

4. **Read the result.** The helper prints:
   - the session ID;
   - the prompt path;
   - the raw JSON output path;
   - the extracted Claude text if it can parse it.

5. **Act on valid findings.** If Claude finds a real issue and the user asked
   for implementation, fix it. If Claude is wrong or speculative, do not make
   the change; briefly document why.

6. **Report high-signal results.** Summarize Claude's actionable findings to the
   user. Do not paste long raw output unless the user asks for it.

## Session Handling

The helper script uses `claude -p --session-id <uuid>` for a new review thread
and `claude -p --resume <uuid>` for follow-up turns.

**Default policy: maintain one continuous review session.** The helper resumes
the stored session id on every run. Keep doing that — review after review goes
into the same thread.

- **Do NOT pass `--new-session` on your own judgment.** Not for a "different
  topic," not for a new experiment, not for a new issue. The user decides when

From shannonshell/shannon, MIT.

More from shannonshell/shannon